scala/scala

Discord is now Scala’s main chat platform. Please join us at https://discord.com/invite/scala

Luis Miguel Mejía Suárez

@BalmungSan

Oh right, I always forgot about export

econoraptor

@econoraptor

Hello, I have a few questions about typing and isInstanceOf. For starters, what's going on here when this returns true?

val x: AnyVal = 3
x.isInstanceOf[Int]

econoraptor

@econoraptor

aiui anyval gets precompiled to concrete primitives, which would explain this, but not another question i have, so I want to confirm that's indeed what's happening

Rob Norris

@tpolecat

3 is an Int literal so it compiles to an Int. You have typed it as AnyVal, which is fine because it’s a more general type.

But we don’t use .isInstanceOf much in Scala. It’s best to think in terms of declared/inferred types, not what’s happening at runtime.

econoraptor

@econoraptor

pattern matches on type use isInstance of though, right?

*isInstanceOf

Martijn

@martijnhoekstra:matrix.org

[m]

For type patterns, yes (e.g. case x: Int)

Rob Norris

@tpolecat

Type patterns do, and must. Structural (ADT) matching also does, but this is an implementation detail. It could work differently.

Martijn

@martijnhoekstra:matrix.org

[m]

For constructor patterns, as an implementation detail, yes, but equivalent to unapply (and its specced as such, there is no observable difference)

econoraptor

@econoraptor

ok, so what about this?

type A <: Int
object A {
  def apply(i: Int): A = i.asInstanceOf[A]
}

val isIntA = 3.isInstanceOf[A]

First, the compiler warns me that the type test for A cannot be checked at runtime so I'm curious about that. Second, it tells me that 3 is an A, which surprised me.

Rob Norris

@tpolecat

A is not a class type.

That should be an error. It will probably tell you “foo” is an A.

Don’t use asInstanceOf, don’t use type patterns.

They are unsound if you ignore the warnings.

And they have deeper problems.

econoraptor

@econoraptor

"foo" is not an A, apparently. But point taken about ignoring warnings. Partly I'm just trying to better understand what's going on internally here since I've been running into enough surprising things that there's clearly some things I don't understand

what's the significance about A not being a class type?

Rob Norris

@tpolecat

It’s an abstract type that is completely erased at runtime. There is no associated class or any other mechanism that would let you tell whether an arbitrary value is an A or not.

The best way to think about this in my opinion is to assume that all types are erased and there is no way to tell what anything is at runtime. This is in fact true in general, but for things with primitive or non-parameterized class types you can look at runtime. But the set of things you can check at runtime is much smaller than the universe of types the compiler understands.

econoraptor

@econoraptor

Ah, I didn't realize that. That definitely helps explain some of the other behavior I was confused by

Rob Norris

@tpolecat

https://typelevel.org/blog/2017/02/13/more-types-than-classes.html

This is a good article but a bit old. Replace “scalaz” with “cats”.

econoraptor

@econoraptor

This looks good. I'll read it and then probably have a couple more questions that I think might be a bit orthogonal. I've watched a couple of your talks and read some of your posts as I've been getting into scala and they've been super helpful, so thx for that as well.

Rob Norris

@tpolecat

Oh cool. Well this is a good place for questions. We’re moving to Discord so you might try that out as well. https://sca.la/discord iirc

Ya.

And https://sca.la/typeleveldiscord for more FP-focused stuff.

econoraptor

@econoraptor

i was aware of the scala discord, but i'll check out the typelevel one too

Luis Miguel Mejía Suárez

@BalmungSan

I guess the time to revive my old ("gamer") discord profile has come.

econoraptor

@econoraptor

ok, this article has convinced me that reasoning about runtime types is difficult

Rob Norris

@tpolecat

It’s easy if you pretend it’s 100% impossible!

Matthew Rooney

@javax-swing

@Woody88 the issue you were facing is that for a Monad ap was being defined in terms of fmap and fmap was being defined in terms of ap. by overriding the definition of fmap in your Monad trait to be defined in terms of flatMap it fixes the issue

     def fmap[B](f : A => B) :M[B] = ma.flatMap(a => pure(f(a)))

Justin Reeves

@justinallenreeves

I guess the time to revive my old ("gamer") discord profile has come.

@BalmungSan
If you have gmail or protonmail, you can use a alias when you signup for an account ${user}@gmail.com => ${user}+${alias}@gmail.com
The aliases still get sent to you, but they have alias in the TO field so you can properly filter via domain
Some thing I wish I knew about 20 years ago for setting up spam filters.

Woodson Delhia

@Woody88

@javax-swing Thank you. It works. I just want to have a better understanding so because Monad trait was defining ap using fmap it causes the loop. I guess that is because Monad trait is reference to the fmap within Applicative. However, why does @main def main = println(List(1).fmap(_ * 2)) refer to the Monad version of instead of directly using the Functor implementation?

Luis Miguel Mejía Suárez

@BalmungSan

@justinallenreeves :+1:

@Woody88 note that you also shouldn't provide the three givens.

You should just provide the instance of the most specific type (in your example Monad)

Woodson Delhia

@Woody88

@BalmungSan Hmm... So, the instance that will always be pass will be the monad instance? What if I have a generalized function that just needs a Functor constraint, how will the compiler find that instance?

For example:

def someGeneralizedFunc [F[_]: Functor](fa: F[String]): F[Int] = fa.fmap(x => 3)

someGeneralizedFunc(List("yo"))  // I don't think this will work with just a Monad instance for List

Luis Miguel Mejía Suárez

@BalmungSan

@Woody88 yes it will find it.

From a theoretical point of view, any type that forms a Monad can form a Functor

Because you can always provide a lawful implementation of map in terms of flatMap and pure

However, in Scala is even simpler than that. Monad[F] extends Functor[F], as such, it is a subtype.

And due Liskov we can always pass a subtype.

Woodson Delhia

@Woody88

@BalmungSan I get that but it's because you Monad directly extends Functor. I have Monad only extends Applicative, and Applicatve extends Functor. I just gave it try and it doesn't work unless I have Monad extends Applicative with Functor.

3 replies

However, isn't it repetitive since Applicative already extends Functor?

Colin Fraizer

@cfraizer

Where can one download the 3.0 API documentation as a tar/zip file?

Swoorup Joshi

@Swoorup

does scala have a dbg! macro as in rust?

heksesang

@heksenlied:matrix.org

[m]

What does this mean when I try to call map on a tuple?

[error]    |                         Found:    Object with PolyFunction {...}
[error]    |                         Required: PolyFunction{apply: [t](x$1: t): Any}

1 reply

Where communities thrive

People

Repo info

Activity