# Nullability and type inference
**URL:** https://discuss.kotlinlang.org/t/nullability-and-type-inference/436
**Category:** Uncategorized
**Created:** [October 16, 2014, 8:41am UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436 "2014-10-16T08:41:41Z")
**Posts on this page:** 7
**Page:** 1
### Author:  [@Oliver\_Plohmann](https://discuss.kotlinlang.org/u/Oliver_Plohmann)
#### Post date: [October 16, 2014, 8:41am UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/1 "2014-10-16T08:41:41Z")
This is something I wanted to ask for a long time. Now it struck my mind again. Type inference works well for non-null vars, e.g.
> val str1 = "foo"
But for optional nulls type inference seems not to work:
> val str2 **?** = "bar" // compiler error val str2 = "bar" **?** // compiler error
Think I tried out all positions where you can put the ?. Seems like I have to declare the type explicitly for things to compile if the var may be null:
> val str: String? = "baz" // compiles fine
While this is not really dramatic, it still feels a bit like a nuisance. Am I doing something wrong or could it be done in the compiler? Just curious ...
Regards, Oliver
P.S.: Thanks for M9. Kotlin is now in a state that you can really develop with it 🙂
---
### Author:  [@abreslav](https://discuss.kotlinlang.org/u/abreslav)
#### Post date: [October 16, 2014, 9:04am UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/2 "2014-10-16T09:04:30Z")
I think you might have got the wrong impression about what "a?.b" is. It is not "call 'b' on 'a?' ", but rather "safe-call 'b' on 'a' ". I.e. '?.' is one operator name "safe call", not two, dot and questionmark. (Just for the record, it's the same for '?:').
> val str2 **?** = "bar" // compiler error val str2 = "bar" **?** // compiler error
So this sytax is not Kotlin :) You can't append question marks to expressions, nor variable declarations. I can think of a possible meaning for the first line, but can't see how to sensibly define semnatics for the secons, so that it is compatible with '?.' semantics.
---
### Author:  [@Oliver\_Plohmann](https://discuss.kotlinlang.org/u/Oliver_Plohmann)
#### Post date: [October 16, 2014, 9:37am UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/3 "2014-10-16T09:37:49Z")
Hello,
what I mean is this: Instead of “val str: String = “baz”” I can say “val str = “baz””. In the same way, in the case of “val str: String **?** = “baz”” I would like to leave the type information away as well (that str, which may also be null as in "val str: String **?** = “baz”, is a String). So this is why I came up with “val str **?** = “baz””. I’m aware of the elvis operator, but there is no method invocation being done here. It is about being able to say “val str: String **?** = “baz”” without having to declare the type, which here is String.
The point of the question was to ask whether there is a way of saying “val str: String **?** = “baz”” without having to declare a type String as in this example and let the compiler do the type inference work. I think this would be really nice. I don’t see why type inference should not be at the developers disposal when declaring a variable that may be null.
> So this sytax is not Kotlin 🙂 You can’t append question marks to expressions, nor variable declarations.
The point was to communicate that I had tried out all kinds of combinations to no avail. Maybe I’m just getting it wrong. I don’t know … 😉
---
### Author:  [@Oliver\_Plohmann](https://discuss.kotlinlang.org/u/Oliver_Plohmann)
#### Post date: [October 16, 2014, 8:01pm UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/4 "2014-10-16T20:01:52Z")
Ha! Got an idea and it works:
> fun \ T.orNull() : T? { val t: T? = this return t }
>
> fun main(args: Array\) {
> var str = “”.orNull()
> str?.toUpperCase()
> str.toUpperCase() // does not compile as desired
>
> var num = 123.orNull()
> var otherNum = num;
> otherNum = otherNum?.inc()
> otherNum = null
> otherNum.inc() // does not compile as desired
> }
I used val in my first post instead of var. Think that’s why it could not be understood. Mea culpa. Anyway extension methods in Kotlin are really amazing :-). Well, I still prefer \. But who cares. Maybe something like orNull() should be in the base library?
---
### Author:  [@abreslav](https://discuss.kotlinlang.org/u/abreslav)
#### Post date: [October 16, 2014, 10:30pm UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/5 "2014-10-16T22:30:14Z")
This is a good solution indeed. Feel free to file an issue for standard library
---
### Author:  [@Oliver\_Plohmann](https://discuss.kotlinlang.org/u/Oliver_Plohmann)
#### Post date: [October 17, 2014, 9:12am UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/6 "2014-10-17T09:12:38Z")
Okay, here it is:
[KT-6052](https://youtrack.jetbrains.com/issue/KT-6052)
---
### Author:  [@abreslav](https://discuss.kotlinlang.org/u/abreslav)
#### Post date: [October 17, 2014, 9:49am UTC](https://discuss.kotlinlang.org/t/nullability-and-type-inference/436/7 "2014-10-17T09:49:29Z")
Thanks