https://kotlinlang.org logo
Channels
Powered by
#stdlib
  • k

    kmruiz

    04/18/2016, 1:18 PM
    This could change because there isn't any contract about that, they could return a LinkedList instead
  • k

    kirillrakhman

    04/18/2016, 1:19 PM
    you can do 
    is RandomAccess
  • k

    kmruiz

    04/18/2016, 1:22 PM
    hmm then is there any way to overload the extension method to any MutableCollection<T> that implements RandomAccess?
  • k

    kmruiz

    04/18/2016, 1:23 PM
    would be something very interesting
  • k

    kirillrakhman

    04/18/2016, 1:23 PM
    no
  • k

    kirillrakhman

    04/18/2016, 1:24 PM
    well yes, but it wouldn't make sense
  • k

    kirillrakhman

    04/18/2016, 1:24 PM
    at compile time, you generally don't know it it implements 
    RandomAccess
  • k

    kmruiz

    04/18/2016, 1:25 PM
    so if we want to optimize for RandomAccess we need to do a typecheck
  • k

    kirillrakhman

    04/18/2016, 1:25 PM
    yes but since the method needs to be inlined, this would produce a lot of unused bytecode
  • k

    kirillrakhman

    04/18/2016, 1:26 PM
    you would have the 
    if else
    check at every call site
  • k

    kmruiz

    04/18/2016, 1:29 PM
    this could be solved in compile time using specialization, if we could "change" the return type of mutableListOf to return a MutableRandomAccessList<T> (or something like that) we could get rid of that bytecode in compiletime
  • k

    kmruiz

    04/18/2016, 1:29 PM
    I don't know currently the cost of something like that, I suppose that very big because it's changing the contract of a stdlib method that is used everywhere
  • k

    kirillrakhman

    04/18/2016, 1:29 PM
    mhm
  • k

    kirillrakhman

    04/18/2016, 1:30 PM
    actually, having intersection types for return types would be ok, too
  • k

    kirillrakhman

    04/18/2016, 1:30 PM
    mutableListOf
    would return 
    MutableList & RandomAccess
  • k

    kirillrakhman

    04/18/2016, 1:30 PM
    and 
    removeFirst
    would accept 
    MutableCollection & RandomAccess
  • k

    kirillrakhman

    04/18/2016, 1:30 PM
    the second part is already possible, but the first is not
  • k

    kmruiz

    04/18/2016, 1:32 PM
    Probably is a thing to ask to the Kotlin guys up there if there is any planning for intersection types in extension methods
  • k

    kirillrakhman

    04/18/2016, 1:33 PM
    the problem is not with extension methods
  • y

    yole

    04/18/2016, 1:33 PM
  • k

    kirillrakhman

    04/18/2016, 1:34 PM
    this works
    Copy code
    inline fun <R, T> R.removeFirstX(predicate: (T) -> Boolean): T? where R : MutableCollection<T>, R : RandomAccess {
   // ....
}
  • k

    kirillrakhman

    04/18/2016, 1:34 PM
    no change necessary
  • k

    kmruiz

    04/18/2016, 1:34 PM
    Thanks @yole
  • k

    kirillrakhman

    04/18/2016, 1:35 PM
    but it's awkward at the call site, because the only way to have something that is both 
    MutableCollection<T>
    and 
    RandomAccess
    is to have a subtype that implements both
  • k

    kirillrakhman

    04/18/2016, 1:36 PM
    e.g. 
    ArrayList
  • k

    kirillrakhman

    04/18/2016, 1:37 PM
    if I'm not mistaken, intersection types are a compile-time feature, they don't exist on the JVM
  • k

    kirillrakhman

    04/18/2016, 1:37 PM
    so you can't declare a variable with type 
    MutableCollection & RandomAccess
  • y

    yole

    04/18/2016, 1:37 PM
    type declarations of local variables don’t exist on the JVM either simple smile
  • k

    kirillrakhman

    04/18/2016, 1:38 PM
    could you elaborate?
  • k

    kirillrakhman

    04/18/2016, 1:38 PM
    I know there is a stack, but you need to know the type of what is on the stack, don't you?
1...646566...149Latest