Channels
#stdlib
- k
kirillrakhman
04/18/2016, 2:01 PMbut you didn't win anything. your variable is still of type
but notMutableList<Int>RandomAccess - k
kirillrakhman
04/18/2016, 2:01 PMif you try to use it as a parameter for a method that expects
, it won't compileRandomAccess - k
kmruiz
04/18/2016, 2:05 PM-.kt - k
kmruiz
04/18/2016, 2:05 PMThe problem here a think it's that the compiler does not infer R nor T, solving this issue it would work fine I think - k
kmruiz
04/18/2016, 2:06 PMHowever this code does not solves anything because it ignores completely the compiler and depends on the client of the library to work - k
kmruiz
04/18/2016, 2:06 PM😞 - k
kirillrakhman
04/18/2016, 2:06 PM
is of typeints
nowArrayList - k
kirillrakhman
04/18/2016, 2:06 PMyou still didn't win anything - k
kmruiz
04/18/2016, 2:07 PMI know, I think it's impossible without touching the compiler to solve this at compile time - k
kirillrakhman
04/18/2016, 2:08 PMthe only way to have intersection types as return types is to let the compiler infer it:
but this was specifically prevented by @dmitry.petrov IIRCfun <T> intersection() = if (true) mutableListOf<T>() else null as RandomAccess - d
dmitry.petrov
04/18/2016, 2:09 PMnot me directly simple smile @stanislav.erokhin has more to tell on this particular subject - k
kirillrakhman
04/18/2016, 2:10 PMah, thanks! - k
kirillrakhman
04/18/2016, 2:11 PMI think you could even make it work on the Kotlin side but Java interop would be a disaster - k
kirillrakhman
04/18/2016, 2:11 PMall you need is to know at compile time that something has both types and cast it on the call site - k
kirillrakhman
04/18/2016, 2:12 PMnot surprisingly, Ceylon has it: http://ceylon-lang.org/documentation/tour/types/ - d
dmitry.petrov
04/18/2016, 2:13 PMYes, that's a long topic of internal debates regarding "smart casts vs rich flow typing" - k
kirillrakhman
04/18/2016, 2:14 PMwas I right assuming that Java interop was a big hurdle? - d
dmitry.petrov
04/18/2016, 2:16 PMYes, that's one of the main reasons. - k
kmruiz
04/18/2016, 2:17 PMThen Kotlin probably will never implement union/intersection types - k
kirillrakhman
04/18/2016, 2:17 PMwell, there are ways - k
kirillrakhman
04/18/2016, 2:18 PM
for example@InlineOnly - k
kirillrakhman
04/18/2016, 2:18 PMyou can't call it from Java => no need for interop - d
dmitry.petrov
04/18/2016, 2:18 PMYes. @YouCantCallItFromJava simple smile😂 2 - k
kirillrakhman
04/18/2016, 2:18 PMeven better 😄 - d
dalexander
04/18/2016, 2:18 PMThat seems like it would solve a lot of problems =p - d
dmitry.petrov
04/18/2016, 2:19 PM... and introduce new ones. Like, "we DON't use @YouCantCallItFromJava, because we have several MLOCs of Java sitting next to us" - k
kirillrakhman
04/18/2016, 2:19 PMalso a compiler flag that makes @YouCantCallItFromJava compiler errors - d
dalexander
04/18/2016, 2:20 PMYeah, of course. If Java -> Kotlin interop was useless, then it wouldn’t be supported at all. (And Kotlin would probably have the same level of uptake that we see in other JVM languages that have made that decision) - k
kirillrakhman
04/18/2016, 2:22 PMbut still, you can't call @InlineOnly from Java so it's not that absurd to imagine - k
kmruiz
04/18/2016, 2:23 PMI think it can be really useful for the stdlib