Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 18:48
    som-snytt commented #12434
  • 18:02

    SethTisue on main

    FAQ: add question about _ in la… (compare)

  • 18:02
    SethTisue closed #2142
  • 18:02

    SethTisue on main

    FAQ: is this StackOverflowError… (compare)

  • 18:02
    SethTisue closed #2143
  • 18:02
    SethTisue commented #2144
  • 18:02

    SethTisue on main

    update JDK compat page (add 18,… (compare)

  • 18:02
    SethTisue closed #2144
  • 18:00

    SethTisue on 2.13.x

    2.13: advance project SHAs (#14… (compare)

  • 18:00
    SethTisue closed #1464
  • 17:35
    SethTisue synchronize #1464
  • 16:43
    SethTisue edited #1464
  • 16:43
    bishabosha edited #9716
  • 16:43
    SethTisue synchronize #1464
  • 16:39
    scala-jenkins milestoned #9716
  • 16:39
    bishabosha assigned #9716
  • 16:39
    bishabosha review_requested #9716
  • 16:39
    bishabosha review_requested #9716
  • 16:39
    bishabosha labeled #9716
  • 16:39
    bishabosha assigned #9716
Vinayak Pathak
@vinayakpathak
or is it the "typeclass trick" using implicits
Fabio Labella
@SystemFw
which is half-broken, and it's going away

or is it the "typeclass trick" using implicits

No, I don't see any relevance to typeclasses and implicits here

Vinayak Pathak
@vinayakpathak
well when you define a typeclass such as trait Monoid[T] you are saying that a Monoid[T] exists for every T however, the way you implement it using implicits, it is guaranteed that it only exists for the T's for which you have defined a typeclass instance
so in some way you are encoding existentials using universals?
Fabio Labella
@SystemFw
no, this is a different thing
Vinayak Pathak
@vinayakpathak
okay
Fabio Labella
@SystemFw
first of all, your example has not much to do with typeclasses or implicits
you can replace Monoid[T] with anything that has a type parameter and explicit instances, and your question remains the same
second, you can look at existentials to be linked to information hiding
as in "I have a Monoid for some type, but I dont know which type"
whereas when you get Monoid[Int], you know that the Monoid is for Int
Vinayak Pathak
@vinayakpathak
true
Fabio Labella
@SystemFw
matter of fact
encoding existentials through universals in a simple way requires higher rank polymorphism (which I've talked about in this channel a few days ago), which scala doesn't have
Vinayak Pathak
@vinayakpathak
i see
Fabio Labella
@SystemFw
you could simulate it in this case, but then again abstract types already give you a way of encoding existential quantification, so...
Vinayak Pathak
@vinayakpathak
i'm thinking about your statement: "I have a Monoid for some type, but I dont know which type"
Fabio Labella
@SystemFw
yeah, that doesn't hold
go back to Model
you have a Model for some State, but you don't know which State
hence when you try to coerce it to Int, type mismatch
Vinayak Pathak
@vinayakpathak
that is if I write val m: Model = new Model {...}?
Fabio Labella
@SystemFw
yes
Vinayak Pathak
@vinayakpathak
i see i see
so existential types allow us to define values that have this information hiding feature
if you wanted to define a function that took a Model without knowing its State, you could do that with universal types too no?
Fabio Labella
@SystemFw
nope
Vinayak Pathak
@vinayakpathak
def f[Model[_]] = ??? ?
Fabio Labella
@SystemFw
because you would have Model[State]
wait
the _ you are using there doesn't mean what you think it means
because scala is weird
like
I know what you mean
but your code there doesn't mean that
Vinayak Pathak
@vinayakpathak
i see!
Fabio Labella
@SystemFw
it means that you are taking a higher-kinded type Model[_]
or in other words
def f[Model[_]] = ???
is not the same
and in fact has nothing to do with
def f(model: Model[_]) = ???
this is what you mean
a wildcard type
which again, it's a feature that encodes existentials
Vinayak Pathak
@vinayakpathak
actually yes, that's what i meant indeed
isn't this saying that f takes a value of type Model but doesn't care what State is
Fabio Labella
@SystemFw
yes
it's introducing an existential again